György Buzsáki and David Poeppel (Archive BI 84)

[00:00:02] Speaker A: Every science in the history, physics, chemistry, genetics, computer science, they came up with their own defined vocabulary. Neuroscience hasn't. Neuroscience is just take it for granted that we are already given a vocabulary and we have a roadmap. All we have to do is take those words and identify their mechanisms. [00:00:22] Speaker B: I'm a naive realist. I believe the world is as it is and we can figure it out. And I would like to use the best, put our best foot forward. Right. And so I think the slightly too casual dismissal of the contribution of the psychological sciences comes at a cost that I don't want to incur. [00:00:42] Speaker A: David, let me push you in a corner now, okay, please. Here is a quote from the things to be explained are human cognitive faculties. So here are my questions. [00:00:56] Speaker B: It's important, I think, that people realize that we can be friends, colleagues, and actually appreciate the work and just try to understand these very difficult problems in debate. There's just nothing wrong with that, that hopefully makes better science, makes our arguments deeper, and we're working on extremely difficult problems. [00:01:21] Speaker C: This is brain inspired. [00:01:34] Speaker D: Hey everyone, it's Paul. Brain Inspired is off this week because of some scheduling changes that the transmitter has needed to make. I will be back next week and then the. And then resume an every two week schedule. But I didn't want to leave you hanging for a week here, so I decided to reach way back in the archive from 2020, I believe, so six years ago. This is Yuri Buzaki and David Popel. I had them on because they had both written opinion pieces in Enuro in sort of a debate kind of style. And so this is off the heels of Yuri Buzaki's book, the Brain from Inside Out. So Yuri believes it's time that we use what we know and what we continue to discover about brain mechanisms to better understand the psychological concepts that we referred to as explanation for minds. Essentially he wants to throw away psychology or the names that we use in psychology and focus on brain mechanisms to sort of rebuild a more valid set of concepts in psychology. And David believes that the psychological concepts are fine, they're constantly being refined as is, but they're just as valid as objects of study to understand minds. Okay, so that's essentially what this conversation between the two is about. Also there are some special guest questions from Paul Ciesek and, and Jon Krakauer. If you have heard this before, I hope that you enjoyed this blast from the past. If you're relatively newer as a listener to burn inspired, I hope that this will be a nice treat for you. In the interim before I Am back. Okay, here are Yuri and David Yuri, your recent book the Brain from Inside out, which I love and I also loved Rhythms of the Brain. They're both written really, really well and it's just something for authors to aspire to. One of the wonderful features in both those books are the summaries that you give at the end, the little chapter summaries. They're so good. When I was in graduate school, after I read Rhythms of the Brain, I took the book which the library had to the copier which also had a scanner. And at the time this seemed pretty fancy to me. And I scanned in all of the summaries and I still have it on my computer. So it's made it through a few different computers now. And just recently I just pulled it up actually to make sure I, I still had it. So nice job on both books. But the Brain from Inside out is also filled with a bunch of quotes. So I'm going to start with what I think is a relevant quote for today and see if you guys can guess who it's from. Quote. I am convinced that an important stage of human thought will have been reached when the physiological and the psychological, the objective and the subjective are actually united. When the tormenting conflicts or contradictions between my consciousness and my body will have been factually resolved or discarded. Can you guys guess who that is? [00:04:47] Speaker B: No, but I would have said Descartes. [00:04:49] Speaker D: Ooh, Descartes. Pretty good guess. [00:04:51] Speaker A: Yuri, any guess, somebody like Descartes could say that. [00:04:55] Speaker D: Well, Pavlov would be. I think he would take that as a compliment. He was guessed as Descartes. Yeah. [00:05:01] Speaker A: There is nothing new under the sun. There is nothing new under the sun. [00:05:07] Speaker D: That's right. It's an old. Yeah, really it's. And Descartes is. It could very well have been Descartes. So people can be very opinionated about this topic. So I'm confident today that we will resolve it, I'm sure. Okay, here's another quote before we really get the ball rolling. I analyzed how an undefined and unagreed upon terminology which we inherited from our pre neuroscience ancestors and never questioned, has become a roadblock to progress. Yuri, any guesses? [00:05:38] Speaker A: Could be Case van der Wolf. [00:05:39] Speaker D: Descartes. [00:05:40] Speaker B: Perhaps I would have guessed Yuri. That's Yuri. [00:05:44] Speaker D: There you go. Yeah, that's Yuri. [00:05:47] Speaker B: That's a tweet that Yuri wrote for sure. [00:05:49] Speaker D: That's a tweet. Yeah. [00:05:51] Speaker A: But you know, I can come up with several people. Of course, you, you know, again, thoughts mature very slowly and they get embedded into the right environment and that that process is very slow. [00:06:06] Speaker D: Yeah, well, okay guys, I'm going to ask you both to summarize, you know, two minutes, you know, whatever it really takes. Not, not more than 45 minutes, kind of just briefly summarize positions your perspectives on these issues and then we'll move on from there. And Yuri, let's just start with you because the way that the articles worked is you posted your 20 year old article, which David then got to respond to. And then after you guys summarize, then we'll come back and I'll let Yuri also respond to David's response. So Yuri, take it away. [00:06:41] Speaker A: Well, historically, research on the brain has been working its way in from the outside world, hoping that such systematic exploration will take us someday to the middle and on through the middle to the output. Ever since the time of Aristotle, philosophers and scientists assumed that the brain, or more precisely back then the mind, is initially a blank slate filled up gradually with experience in an outside in manner, an alternative, a brain centric view. The one I'm proposing is that self organized brain networks induce a vast repertoire of preformed neuronal patterns while interacting with the world. Some of these initially nonsensical patterns acquire behavioral significance, or you can call them meaning. Thus, experience is a process of matching pre existing neural dynamics to events of the world. The perpetually active internal dynamic is the source of cognition, a neural operation disengaged from immediate senses. Accordingly, in my book that you kindly mentioned, the Brain from Inside Out, I discuss three major first, how we got to our presently dominant framework in neuroscience. The second part argues in favor of the idea that all sources of knowledge, including our perceptions, memories and plans, derives from actions. In the third part, I illustrate the advantages of self organized pre configured brain and present it as an alternative to the blank slate framework now thinking about. Part of the book began 20 years ago when I wrote that review, a manifesto if you want, about the origin of our scientific terminology, which is the main topic of the book's introduction. In this old review, I summarize how neuroscientists began to study the brain, buying into a system created by philosophers and psychologists for understanding the soul and the mind. Without ever asking how those terms, whose brain function neuroscientists are trying to understand, such as consciousness, attention and so on, were brought into our thinking in the first place. I argued that this outside in framework may not be the best strategy to understand the brain, although back then I didn't have enough ammunition to offer an alternative. The arguments laid down in that Unpublished manuscript have become more popular over the years. So to test the waters, I resubmitted the paper and got four good reviews. Now publish. This text with its 20 year old references and thinking is a good target for contemporary cognitive scientists like David Poppel, who can rip it apart and offer better substitutes. [00:09:18] Speaker D: All right, that's good introduction, David. You want to just take it from there? [00:09:23] Speaker B: Sure, let me give. So there's a few things for context I want to start with. First of all, like you said, Paul, and for me as well, Yuri's book Rhythms of the Brain is absolutely foundational and super important and required reading in my lab. It's an absolutely kind of masterpiece. And I have to say, very kindly. I once saw Yuri at a lecture I gave. I think it was a Rutgers, and he signed the book. I had the book with me and he signed it saying all intelligent people think about timing, which I thought was very good. Thanks for that, Yuri. And then of course, I voraciously read his recent book. Now there's a kind of larger point I want to start with before we debate all kinds of nitty gritty argumentation. That is, I want to make clear to listeners, in particular students and postdocs or trainees, how useful it is to be able to have debates, even among friends and colleagues. So a few years ago I was teaching a class at Cold Spring harbor, and in that context I got into a really intensive debate with one of my very close friends. And we really had it out for an hour or so and the students came back and felt uncomfortable about this level of back and forth, which I found really disturbing because it has nothing to do with our regard for each other, with our respect for each other's work and so on. We just, we were disagreeing on some points and we really went after each other and this, the trainees didn't like that. And I found that not good, actually. It really. I found it bewildering and disturbing. And one of the things I really appreciate about being able to sort of bounce back and forth ideas with Yuri and having a discussion today is that it's important, I think that people realize that we can be friends, colleagues, and actually appreciate the work and just try to understand these very difficult problems in debate. There's just nothing wrong with that, that hopefully makes better science, makes our arguments deeper, and we're working on extremely difficult problems. And so for. Don't begrudge us that we can like each other and still debate points because we're trying to move the thing forward. I found that a really odd reaction a few years ago and it really kind of moved me. So I'm very happy that we can do something about the topics. [00:11:43] Speaker A: So I had exactly the same memory and experience as your students many years ago when my mentor Andrew Grashyaan invited Walter Freeman, our Walter Freeman, to give a talk in Hungary. And then during his talk, Grazchan attacked him. And it was strange to me, you know why he invites a person from faraway lands to attack him. And at lunch Walter Freeman said, oh, I never had such a good time for a long time. [00:12:14] Speaker D: But it made you uncomfortable, Giri at the time? [00:12:17] Speaker A: Yeah, well, it. Yeah, just like David's students. But I think David's students also realized in time that discussion is the way how to move forward. [00:12:27] Speaker B: Exactly. [00:12:27] Speaker A: If you don't play out your cards, how do you know what's in the cards? [00:12:32] Speaker D: Well, I was going to ask David before you told that story whether he thought that it was a generational thing or whether it was a developmental aspect that happens every generation. I really don't. Is it snowflakes or is it a normal developmental trajectory? [00:12:48] Speaker B: I must say, I don't know. So the paper that I wrote just in reaction to Yuri's paper that Christophe Bernard kindly invited us to comment on, is a graduate student. So he's a young Federico Adolfi. He's a young scientist who works on computational topics and psychophysics. He just has a big appetite for these topics and knowledge and likes to read. So it can't be purely generational. I do remember in graduate school, growing up in a context where debate was vigorous, maybe sometimes too vigorous and valued, and I think that's not always the case so much. I think it's a risk that we take to be too. I mean, these are not things that are ad hominem. We can say, I just think you're wrong about this argument and we try to move the needle and I think people should accept that as part of a positive part of the scientific process, actually. [00:13:35] Speaker D: Is that why you so in your article, and I'm sorry, it's already a digression, which is totally fine because that's what we should do. But is that why you are so complementary throughout the article of Yuri's work? Just to ward off the potential? [00:13:49] Speaker B: Well, I think it's important to be clear because. So this is not about. So for instance, take the last part of. Or I think it's the last part of Ury's book where there's tons of places where of course, I agree those are very fundamental points in Particular the last. [00:14:03] Speaker D: Not good for the show, David. That's not good for the show. [00:14:05] Speaker B: No, don't worry, I'll disagree soon. So there's a part of the book, for instance, the notion of preformed self organized dynamics that form the basis. I'm absolutely on board with that. I think that's actually where I think it would be fun to work together on how to move that forward. I completely agree, but the point is one can agree on certain things, on data, on how to do experiments, and disagree about the foundations of how we move this forward. And that is where I disagree. And that's sort of the things I tried to highlight in our reaction to Yuri's piece. [00:14:39] Speaker D: You want to summarize what the. [00:14:40] Speaker B: So I'll summarize quickly. So, I mean, maybe the easiest way to summarize it is that we stand in defense of the outside in view rather than the clear inside out view that Yuri outlines, because we think that it happens anyway. So that the inside out, the kind of stringently, stridently inside out view that Yuri argues for in the book is not really what happens. You have implicit hypotheses, you're an implicit philosopher. You make assumptions that sort of underpin what you're trying to explain. Then you do the very detailed implementational work and then you go back and sort of revise and try to make it very explicit. So we have a sort of slightly cheeky phrase called the implementation sandwich. You can't just start with the wetware or the brain. You actually make some assumptions about what you're trying to explain. Then you do the implementation level work, and then you refine the kind of psychology or cognitive science, but you sandwich the implementation work. You don't just go from a characterization of the hardware to a putative function. And that's where I think we probably disagree. [00:15:49] Speaker D: Okay, so Yuri, I want to give you a chance to respond here to David's perspective. First of all, is that an accurate portrayal of your view or would you say that maybe he mischaracterized some of your goals? Is there a straw man that David is addressing that isn't your objective? [00:16:12] Speaker A: Well, first of all, I'd like to thank David and Federico for taking their time reading my review and parts of my book and writing a beautiful piece expressing differences of opinion. I think this is what we expected. I'm talking about neuroscience and the brain. I think David is representing the psychology and the mindset. He claims that the object of neuroscience is the hardware wetware itself. This is what he has written, but in reality, no Self respecting neuroscientist would agree with such correspondence. The focus of neuroscience is how behavior and cognition are generated by the brain, including by its interaction with its niche, which is the body it supports and other brain it communicates. Other brains it communicates with. I agree with David that it may be premature to abandon the traditional outside in strategy in some areas of cognitive neuroscience because currently there may be no substitute. But my argument is that the time is right for neuroscience because the outside in framework has reached an asymptote in many areas and that process began perhaps 20 years ago. I claim that an inevitable consequence of the outside in framework is viewing the brain as a blank slate. Nobody believes in it, but in practice this idea just does not go away. AI is talking about brain inspired algorithms, but as a result of taking the current state of neuroscience, virtually all AI is based on the tabulo Rasa philosophy. Now David and Federico are trying to push me in a corner of a boxing ring so they can have a vantage point to make their statements. In the tradition of classic cognitive science, they first come up with labels which in their view characterize my work, such as epistemological primacy of the hardware. I like that. The radical implementationist. I like that too. But when these terms are explained in detail in the paper, David, I feel that these texts actually are directed to or towards Henry Markram's original human brain project, which recommends to build up a brain from bottom up. But I do not belong to that camp at all. And the inside out view is very different from the bottom up reductionist formulation. In fact, a good part of my book is trying to explain why such programs have limitations, using similar arguments to David and Federico. Again, the inside out framework has little to do with the bottom up implementationism. It suggests that the debate between the cognitive terms, which are made up by our predecessors, and brain mechanisms, brain mechanisms should be the arbitrators, not the other way around. That's all I wanted to say in response for now. [00:19:01] Speaker D: All right, good. [00:19:02] Speaker B: Well, can I just. Since we're just on this particular issue because. So that it's important, I think that, yeah, the question is how brain mechanisms, you know, brain mechanisms are the arbitrators of what. So in the end our real disagreement is about the utility of the concepts from the psychological and cognitive sciences and how neural data on its own could adjudicate between alternatives that are posed that way. And I think that's so. I'm more optimistic on the side that cognitive science and computation, some amalgam, actually have as an equal Status in terms of evidence as neuroscience. And I think that's where we may disagree actually. That's why we call it epistemological primacy or whatever the phrase was we used to say. Actually in the end, the most serious arbitration comes from the neurobiological data, not from the cognitive science or psychological data. [00:19:55] Speaker A: Yeah. So you ask, what do the brain mechanisms will refer to? Well, what they refer to is already to our preconceived ideas, but they help us to systematize, to abandon some assumptions and reinforce others. I never ever thought seriously, but I realized that perhaps it's easy to misunderstand what I said. I don't suggest that we go into the brain, we do a lot of things there and voila, the cognitive terms will come out. No, we go into anything, into anything. Any thought has a background, it has a context. So we always go with, with something in it. And I say it very clearly in the last two chapters, that the brain always guesses there is nothing new in the world for the brain. Everything is familiar. And then the familiar becomes with time different. If the organism has problems with understanding that. So I understand that part of my brain centered obsession can be viewed as a [00:21:15] Speaker D: radical implementationalist, naive view. [00:21:17] Speaker A: How to. Yes, how to you go here and from elements you build up the brain. But no, I doubt that's possible. And I think this is a platform that we more agree than disagree. There are other areas we will discuss later where there is a disagreement. But I think this is perhaps my mistake to not clarify it enough. And I think the main problem is that indeed that target article probably had the flavor of this, not necessarily the book, but the article was written 20 years ago. It's not a good excuse, but. [00:22:01] Speaker D: David, you were going to respond. Do you want to briefly respond? Yeah. [00:22:04] Speaker B: So I think Yuri raises an important point which is I think where we. I want to make a distinction between implementation driven versus sort of radical reductionism. Those seem to me a little different. So what I want to impute to Yuri is that he is an implementationalist in the sense of anti Marr or something like that. That's not correct about marr. We'll get to that later. But that is the. There is really a higher status of the evidence of the implementation level. That doesn't have to be reductionist in the sense of going to, you know, synaptic mechanisms or quantum mechanics. It has to some level of description. That is the level of the implementation of the putative mechanism is higher ranked in terms of what it is as evidence than some other evidence. And I don't think, I guess evidence is just evidence. And there's no sort of. This is the best evidence, the second best, the third best. I just don't see that. So I absolutely agree with Yuri that some of the more trenchant critique is directed at the research program that kind of Henry Markram pushed, which I also see as not even coherent, actually not even clear how that could go anywhere at all. So I have no sympathy for that, and I wouldn't take the time to write a paper about it. It's just. It's incoherent. [00:23:25] Speaker D: Okay. All right, so this is a great beginning here. And I'm just going to. Right off the bat, we're going to get into questions from my little surprise guests here. And you guys didn't know who these were, and they didn't know. They knew who each other were, but they didn't know what each other's question question was. Okay, so the first question is from Jon Krakauer. And this question will be directed more toward Yuri, but we can, you know, you can both discuss the second. And then before you actually react to that, I'm going to go ahead and play the second question, and that is from Paul Chiesek, and that will be directed more toward David. And then. And then I'll play the two questions, and then we can move on. So here, if you guys are ready, here's Jon Krakauer. [00:24:09] Speaker C: Hi, Yuri. This is John. Sorry that I can't be asking this question in person. I very much look back fondly on our debate at Columbia several years ago. So here's my question. On page 225 of your wonderful new book. You say this about tools. They had to be imagined before their maker could start working. Artifacts are externalized versions of a thought, a reflection of contemplation. Artifacts are semantic entities. So given that you want to get rid of philosophical and psychological language, I challenge you to restate that sentence with only neural implementational language. So no psychological terms allowed. Remove the words imagination, reflection, thought, contemplation, and semantics, and say it all in terms of neurons and circuits. I would contend you won't be able to do it. And please note, you use those psychological terms to hypothesize yourself about externalization. So they are doing work for you conceptually. You may answer that we need to find a new language based on implementational level details, but that's never going to happen. In my view, you're always going to have compressed psychological concepts to express ideas about cognition and neural Data will simply be confirmatory. So my question is, do you really think that you're going to be able to forego the very language that you yourself use throughout your book to conceptualize with? Cheers. [00:25:36] Speaker D: Okay, so that's John's. Can you hold that in mind? And we'll play Paul's. Just because they're of a similar nature but directed in different directions. Okay, so here is Paul. [00:25:48] Speaker E: So my question is for Dr. Popol. In principle, I agree that neuroscience needs behavior and that what we should seek is alignment between theories about low level neural mechanisms and high level psychological concepts. However, the need for having high level concepts does not imply that the particular high level concepts we currently have are necessarily the right ones. And here I agree with Dr. Buzzaki that many of the current concepts of mainstream cognitive neuroscience are inherited from folk psychology and from largely outdated pre scientific ideas about the human mind. In fact, Russ Poldrak has pointed out that about 80% of psychological terms in use today, we're already around in the year 1800, which is long before psychology was actually a science. So despite all we've learned about neural mechanisms in all that time, we still use many of the same old high level conceptual categories to describe behavior. We've subdivided them into smaller categories, but the general taxonomy has resisted change. So my question is, do you believe that the high level concepts have resisted change because they are so good, or is it because they are the words we use to talk to each other about behavior and thus they have a kind of stranglehold on our scientific conversations? If the latter, then what can we do to break that stranglehold? Should we continue to subdivide and modify the same concepts? Or as I would argue, And I think Dr. Buzzaki would argue, should we instead seek a different set of high level concepts that are better informed by all that we've learned about neural mechanisms and real animal behavior? [00:27:25] Speaker D: Okay, so Yuri, you want to respond to David's challenge? [00:27:29] Speaker B: John's challenge? [00:27:30] Speaker A: John's. [00:27:31] Speaker D: Oh, yeah, yeah, sorry, I spoke to him. [00:27:33] Speaker B: There's a David Krakauer too. [00:27:35] Speaker D: I know I've spoken to too many of them recently. And so. Yeah, yeah. To John's challenge. [00:27:40] Speaker A: Yes, John, thank you very much for the question. My response is very similar to what I just gave to David, that I'd like to apologize if you thought that somehow I naively think that we should have opened up the skull, you know, 2,000 years ago and instead of going to the Forum and the Agora in Athens and discuss these things and just look something, some kind of mechanism in the brain and build from there from the details of physiological observations. Then I agree with you. That would be a naive things. I fully agree with you in general as well as with a beautiful paper that you have written with David and several others, that behavior is a fundamental thing. In fact, you may remember, and David probably remembers that in the good old days when you did any kind of psychological experiment with evoke responses, for example, it was mandatory to record the heart rate, respiration, the galvanic skin reflex and so on, because there are so many hidden variables that might fool the investigator that the bold signal change is not a cognitive correlate but just a change in respiration. So I'm fully with you and I agree that whenever we go into doing an experiment we already have our preconceived ideas and these preconceived ideas are coming from other brains. That's. This is why I said, and David picked up this sentence that it's impossible to find nothing in the brain. Yes, it's impossible to find nothing because we have already tons and tons of alternatives which we can exclude. This is what I call the null hypothesis that we go in, we do something and then we can reduce the realm of possibilities we can abandon one at a time. And this is what called scientific progress. So I don't want to be put into the. In the bottom up build, build it up from elements camp because I know it is impossible to break apart a complex system and put it together from its ingredients without understanding the higher rules that complex keep together a complex system [00:30:09] Speaker D: and we could just jump from there. David, do you want to answer Paul's question? [00:30:14] Speaker B: Yes, thank you Paul for that extremely interesting point and of course very much related to points that Yuri makes in his original paper and of course in the recent book Brain from Inside Out. And also actually a point that the editor of the special issue, Christophe Bernhard, makes about naming things and giving them kind of medieval nominalism, giving it, you know, reifying concepts by giving them a name. I think that I am much more optimistic than you in your question and Yuri in his book about the progress that the psychological and cognitive scientists can make and have made. And I think it's very important to make a distinction between the casual everyday use of terminology that has turned out to be quite useful for, you know, talking to each other and the scientific decomposition, the standard. There's a philosophical tradition, Wilfred Sellars, the philosopher makes this very clearly called the manifest image and the scientific image in that particular line of philosophy of science, which is we use certain terminology Just in our, let's say daily locution. Because it works. Right. So Yuri actually gives the example of the concept greed because it's something we understand, it's very easy and we explain behavior quite effectively to each other. It doesn't follow from that that a very systematic decomposition cannot yield new taxonomies or new structures of these concepts. So I think we just like neuroscience, cognitive science or psychology are not that old and we need to give the field a little bit of credit. I mean, I'm a pretty nasty critic generally of the. Of my own field, or at least I'm on record as being pretty nasty to my own field. But I also want to defend that the cognitive sciences in particular, as they're becoming more and more computationally explicit, have made serious progress in decomposing things. We don't just use a concept like imagination or we use a concept like. Well, maybe part of imagination draws on memory mechanisms which themselves are complicated and internal forward models and predictive coding and ensembles. So I think you have to give a little bit more credit to the conceptual analysis that the cognitive sciences and philosophy are offering right now. And be optimistic because there's really. We're not just working with terminology like my personal favorites from the time of Gal, things like relationship to your parents was a conceptual primitive. [00:32:46] Speaker D: You're talking about phrenology time. [00:32:47] Speaker B: Yeah. So you know, you think now retrospectively we make fun of that, but in the context of Gall's, you know, what was going on in the history of how minds were explained, that's a totally reasonable concept. [00:33:02] Speaker D: Right. [00:33:02] Speaker B: The notion that you would have a special attachment to your caregivers. That's not crazy. Those are just. That was. But now we have of course much more richly structured and fine grained and pulled apart conceptual analysis. And I think so I'm actually optimistic. The question that you raise at the end of your question is whether an approach that's different and maybe more endorsed by URI would yield a different taxonomy. That's a totally reasonable question and an empirical question. Right. So as we decompose these things more find successful linking hypothesis to neurobiological mechanisms. We might end up with a slightly different parts list of the mind which would be I think everyone is actually open to that doesn't mean I shouldn't use words like greed anymore or imagination or will. Those are words that we use. They actually do a lot of work for us. I suggest, you know, the work of the philosopher Jerry Fodor explains very well how simple concepts like that they work for daily life but they don't work for scientific analysis. And as long as we keep that separate, I think we're on safe ground. [00:34:01] Speaker D: So it does seem like a lot of this is. First of all, thanks for answering these surprise questions. [00:34:07] Speaker B: Thank you for having surprise questions. [00:34:09] Speaker A: A very good idea. [00:34:10] Speaker B: Very nice idea. [00:34:11] Speaker A: And from good people I really respect. [00:34:15] Speaker B: Yeah, absolutely. [00:34:16] Speaker D: They're okay. [00:34:16] Speaker B: Yeah, they're okay. [00:34:20] Speaker D: So it just seems like so much of this is wrapped up in the language that we use. Yuri, one of the things that you state that you would, I believe you explicitly state it is that you would like to develop from neuroscience sort of a new language to describe the principles and patterns found in circuits and activity, dynamics of neurons. First of all, is that correct? And if so, how's the new dictionary coming along? [00:34:52] Speaker A: It's a nice question. And this is what everybody asks now, how we make progress. Well, David mentioned and we ridiculed Gaul, and it's easy because we are a little bit smarter, collectively speaking as a group. But he was probably smarter than any of us back then. Now, I can say the same thing about. [00:35:18] Speaker B: I absolutely agree. I'm officially on record as agreeing. [00:35:23] Speaker A: Indeed. You know, I think recently we discussed it with Jim Magog, is that, you know, how easy it is to ridicule somebody, and he's very strongly in defense of Franz Gore. Now, the same can be said about the giant of psychology, you know, William James. And the reason why I pick on him is not because I don't like him, but because he's the best. And indeed, his is what I call the James list is something that looked like the top of the top back then in 1890, but today, many of those things look a little funny. And so how do you know, or how do we know that those things that we take seriously today. Let me pick on this term, decision making, a very popular one today, you know, will not be ridiculed 10 years from now or 50 years from now. So how do I imagine this progress? Let me just give you two examples. The first one is that memory, planning, and imagination, these are so different concepts. There are different chapters dedicated to understand and explain it to students in textbooks. There are different people who are doing this research separately, people who are doing imagination. They don't talk or don't necessarily have much to do with memory or planning. But over the past decade or so, starting perhaps with cognitive scientists or starting with us neurophysiologists, we realized that, you know, one thing can be called prediction, the other one can be called postdiction, and one belongs to the past. The other one belongs to the future, related to the current presence. But now, when we look at it, for example, through hippocampal theta oscillations or hippocampal shock wave ripples, when the past, the present and the future just confluences into one entity, then all of a sudden you may wonder whether these separations are justifiable. And then Schachter and then many others who started looking at the imaging of the brain, they were all of a sudden surprised that the same structures, and I could add perhaps the same exact mechanisms that are mobilized when you are traveling back to the past are the same as when you are thinking about the future or making plans. So based on neural mechanism, I can imagine, maybe premature, but these three words in fact will be pull together, just like in relativism of physics, that they are not separate, they are the same. I hope you understand what I'm trying to say. Now there is another interesting concept that I attacked several years ago and then there is a recent. Actually I think Paul Cizak was also one of them. It's about attention. And you know, attention is such a vague, horrible, terrible concept, but it is so useful. And if you look at it slightly differently from the brain mechanism's point of view and you say, aha, it's nothing else but gain. And then David Anderson and Richard Anderson and several other people have produced beautiful results. That indeed is a gain mechanism that we can now explained with a little bit of acetylcholine and inhibition. Or another concept or another territory is running speed. What does running speed have to do with attention? I claim that they are the same thing. The attention is internalized running speed. Running speed is a gain mechanism and the animal goes faster and it produces a gain through perhaps acetylcholine and through the same interneurons that the brain internalizes this initially environment dependent or body dependent function and releases the same acetylcholine to use or produce the same thing such as gain modulation. And that's what the psychologists call attention. So this is another example, I think, when we can make inroads, taking lumping together ideas and perhaps you can come up with a different formulation. This is what happened in physics. Atom is not dividable, but we still call it at home, but with the new understanding, because now at home is better defined. [00:40:16] Speaker D: But, but you mentioned needing new words. Do we need new words? I mean, does you don't need new words? Is it a reckoning of psychological terms that we need or a discarding of them? [00:40:25] Speaker A: You know, either way it worked in in the history, you know, genetics, you know, we are still talking about the same genes, that the words haven't been different, but the meaning, the definition have changed. Every science, every science in the history, physics, chemistry, genetics, computer science, they came up with their own defined vocabulary. Neuroscience hasn't. Neuroscience is just take it for granted that we are already given a vocabulary and we, we have a roadmap. All we have to do is take those words and identify their mechanisms. When I go to give a talk somewhere and I ask and I show James List, then everybody in the audience, without an exception, can identify with one of himself or herself with one of those terms. Yeah, I'm going to dedicate my entire life to figure out attention or I'm going to work on memory, I'm going to work on emotion and so on, as if those entities are really the ones that you will identify when you are starting poking the brain. [00:41:32] Speaker B: But let me ask you, can I ask you so, but I mean, that amplifies my concern about being carefully decomposing concepts into their constituent parts. Just like you raised the issue of atom. Right. We take it as a primitive, but really there are smaller primitives inside there. But in the case of neuroscience, where we do, of course, neuroscience does have a vocabulary for the neurobiological parts list. Do you think we have that correctly identified? So, I mean, you criticize often, rightly, that the cognitive sciences and psychology are kind of lame or using old vocabulary, but that makes a sort of implicit presupposition that neuroscience has its parts list correctly identified. And I'm not so sure that that's true. [00:42:15] Speaker D: Things like action potential. [00:42:17] Speaker B: Yeah, take that. I mean, so we make assumption that, let's say, you know, the neuron doctrine, very successful idea. Is that really the representational or computational primitive in a brain, maybe it's going to turn out in 10 years that it's some weird ensemble that we don't yet know. Maybe it's seven neurons wired up a certain way. So we have a vocabulary for neurobiology, but should we assume that it is actually more successful? I'm not, I'm not so sure that I'm willing to buy that. [00:42:45] Speaker A: You're right, that this argument can go from every level down to a deeper and deeper level. But it's always the case that in order to understand one level, you have to understand something about the interaction of an upper and a lower level. I'm not saying. And you try to put me in a corner and say that. I'm trying to say. I'm not saying that you need quantum physics to explain the knee jerk reflex. Yeah, the knee jerk reflex can be explained the way how I explain it to my mother in law. But we have to ground our observations with other observations which seem to be independent because otherwise we would be double dipping. So when you say we decompose and you know, cognitive neuroscience made enormous progress and you cite your own area, which I, I really respect and I will come back to that, you know, by decomposing the, you know, you just cannot go and say, oh, I'm going to the brain to figure out language. No, you have to decompose those things. [00:43:52] Speaker B: It's just like attention. These are complicated concepts with internal structures. [00:43:57] Speaker A: But then it reminds me a little bit of the program of Hubel and Wiesel, because Hubel and Wiesel said, you know, vision is a complicated thing. Let's break it down to simple ingredients. Let's just study the impact of horizontal and vertical lines on neurons, and then edges and then contours and maybe a little color. And someday from these elementary things that we identified, we'll put together the composition. And of course it didn't work. And I always wondered how many more monkeys, cats, and how many neuroscientists are needed and how many more experiments to move forward. And the answer is that if the entire world, entire 60,000 neuroscientists in the world will do nothing, just the same kind of experiment in the same outside in framework, then we wouldn't get much smarter. So I wonder if that decomposition that is done without grounding is more than just an exercise. And my answer is yes, it's just an exercise. And your answer is that what I recommend is going to the deep air is also an exercise. And this is what science is, you know, it's an exercise. And we'll go to the sandwich problem a little bit later. But this is exactly what we do. We are trying to ground a unknown or less familiar things with a more familiar thing. [00:45:26] Speaker B: But I mean, look, so I respectfully disagree that the decomposition is just ad hoc. I think the point of it is that you take some idea, you know, whatever it is, attention, language, pick your favorite psychological domain, we pick it apart, you know, like peel the onion. Because the assumption is that one of those sub things is a more plausible way to link to the neurobiological thing, right? So if you say, what's the brain basis of languages? I have no idea. It's made up out of dozens and dozens of subroutines. But if I can make it into something that could be Realistically probed at the level of a cell, a circuit, an ensemble, a column, God knows what. Don't you think that that would actually then be at the granularity of linking hypotheses between the neurobiological infrastructure and some part of the concept that we're trying to get? So it's not an ad hoc decomposition, it's a decomposition with a motivation to find, to identify a linking hypothesis to the neurobiology. I mean, I think that's the premise of the research program. [00:46:30] Speaker A: Well, I think what the inside out framework suggests is that it's on a safer ground when we compare these decomposed ideas and comfort them against brain mechanisms than without brain mechanisms. And I can ask the question from you that your composition or decomposition help you to ground your things by relating them to brain rhythms? [00:47:01] Speaker B: No, that's exactly. [00:47:04] Speaker A: But let me remind you that we went out and figured out and tried to work on the system of brain rhythms and their cross frequency coupling without language in mind. So we have observed that these are regularities that are there probably to serve something important. And we can talk about later if you want, why they are important and general principles. And it happened to be a fertile ground for language. Right. [00:47:39] Speaker B: This is a perfect example. So this is the. [00:47:42] Speaker D: David, can you just actually just summarize like that linkage before you know, the phoneme. [00:47:50] Speaker B: So we can take. So, so there's an aspect of. Language comprehension is a complicated problem. It's complicated from a point of view because you have a more or less continuous signal. So as we're having a conversation, there's a physical waveform, it comes to you. And in the end, the remarkable thing of communication is that I send you a physical signal to your ear and it ends up as an idea in your mind, which is cool. The fact that we can have a conversation is actually unbelievably impressive from a brain's eye view, from a mind's eye view. One of the problems you have to do is you have to actually has been known for a long time, certainly, you know, for probably 100 years. It's been discussed, it's certainly since the late 40s. How do you actually chunk the information that's coming continuously into usable bits of information that you can then work with? This has been a pain in the ass of a problem for a long, long time. And some years ago, you know, there were now here functionally driven questions, right? Questions about how do you about this, how would you do this? And a few people, a community of us thought, well, here's a really cool way to do it. So there's this interesting thing of, you know, excitability cycles or oscillations in the brain. What would be if you actually found a link between what they're doing and that particular problem? Because it gives you a potential mechanism for chunking things at different timescales. I mean, and that's a hypothesis that you can chunk things at different levels of sizes, you know, let's say at the syllabic level or smaller levels. And then you can do experiments. Is it real? Is it. You know, some people think it's totally epiphenomenal. I think they're wrong. I think it's real and important. You can begin to really try to parse apart the problem. So there's a case where there's a neurobiological phenomenon which turns out to be immensely useful for querying a functional aspect of language comprehension. But note that it's because we knew the particular problem that had to be solved, namely the problem of segmenting a continuous stream into smaller chunks of different temporal sizes. And that's the case. So the reason I like that line of research is because it's, to me, a really interesting linking hypothesis between a part of language comprehension and a part of neurobiology, both of which are very well motivated. And then you can try to figure out how does it work, how do you manipulate it, you know, how do you actually cash it out? So that's. That's sort of the. And I think that's an area about Yuri and I probably very much agree about that particular issue. That's not. [00:50:18] Speaker A: So, David, what other arguments you think you could use to convince your colleagues that your segmentation that you have done is right? If you don't go and link it to brain rhythms? [00:50:33] Speaker B: There are many. [00:50:34] Speaker A: That's what I call grounding. Because once you start that argument, then you are on safer grounds. [00:50:41] Speaker B: I mean, it's a form of evidence that I happen to find compelling. But of course, let's say you're somebody who works just on automatic speech recognition or some natural language processing. There are psychophysical data that might be as compelling for you. So, for instance, take experiments, psychophysical experiments that my colleague Oded Gitza has done at BU if you take a speed stream and you remove the cues that allow you to have some kind of trunking, then your intelligibility goes away. And if you put in signposts that gives back some kind of rhythmicity in the signal, you get intelligibility back. Right. So those are purely Psychophysical data that are consistent with this. We happen to believe that it's. We want to go to neuroscience because we're interested in neuroscience, but it's. But psychophysical and computational data are of course in my mind, of equal status. You're right that it gives you. The biological data, gives you extra reason to ground it. But these independent forms of data are also legitimate and epistemologically at the same status. I would argue. [00:51:48] Speaker A: David, let me push you in a corner now, okay? [00:51:51] Speaker E: Please. [00:51:52] Speaker A: So here is a quote from you. The things to be explained are human cognitive faculties. So here are my questions. Do we know them? Do we know how many things we need to explain? What authority made up these terms? Isn't there a bit of arrogance by claiming that we know what we are looking for or what you should look for these days? You probably have noticed also that many papers introduce a problem by saying that in order to solve problem X, the brain must have this or that function. Where does this confidence come from? Brain evolution didn't start out generator program where the end product should be the human level cognitive faculties. Instead, brains evolved to induce actions. They learn to predict the consequences of those actions as afforded by a particular environment. The brain is not interested in the true nature of the world. Instead, its main occupation is help its host to survive and prosper its niche. Do we know at all whether attention, motivation and so on, These are entities. So when you look at the brain problem from a wider perspective, then these issues inevitably come up and then we start moving away from the historical aspect of brain research. We started out with the problem of how the mind works and this is what we have to figure out. And we work our way from the top to the bottom. I think Paul Cizak beautifully also explains that, you know, why evolutionary concepts are so important as frameworks to understand that a very complex function cannot be unbroken into simple pieces. But it can be put into an evolutionary context. [00:53:42] Speaker B: Three bears just ran right by my window. Oh my. Black bears. They're right. They're running right by my house. Oh, here comes the. [00:53:51] Speaker D: You're paying attention here. [00:53:52] Speaker B: Holy smokes. [00:53:53] Speaker D: Black bears. [00:53:55] Speaker B: Yeah, well, there's no concept of attention according to Yuri, so I can't see them. [00:53:59] Speaker A: But there is a gain mechanism. And you're gaining. [00:54:02] Speaker B: I'm going to appeal to my game that I saw these. They're right next to my garage. It's incredible. Three cubs that are big and a big mom. Holy cow. Okay, sorry, that was very distracting in my visual periphery. Sorry about. Okay, so I don't feel pushed in the corner at all by that? Because I think it's a fair point. And that's exactly the point that I want to push back on with you, which is why are you confident in our biological primitives? I think that the concepts we use in the cognitive sciences are hypotheses. I mean, so I don't think that there's a particular arrogance to them. I think that there's our hypotheses about the parts list of the mind. Just as the Franz Joseph gall list was a hypothesis apart the parts list of the mind, except that he made some weird extra assumptions about phrenology. Although his organology I think is quite fascinating and good. But those are hypotheses that people I think are willing very likely to abandon when we think about the scientific approach. It's unlikely that people are willing to abandon when we use this terminology just to explain at the superficial level, human behavior. Because like I'm hungry, I'm going to meet my wife for lunch and I'm very greedy, so I'm going to invest an apple or whatever. But the parts list of the mind is a negotiable. And what is the right granularity of analysis of that parts list is negotiable? I think those are empirical hypotheses. They can be tested neurobiologically, which is what we all want to do computationally, psychophysically, theoretically. So there's, I think, a rich reason. So just like I'm claiming that you can't be sure, you know, that maybe the neuron is the right level of analysis, likewise, you're correct, I can't be sure that attention or language or word recognition is the right level of analysis. And I don't think this. I think those are all up for debate. I think everything is up for debate. I'm just wanting to point out that a kind of a decomposition of these concepts to the level of things that could be a linking hypothesis is what we're trying to seek to make progress there. So at the level of, let's say, neural dynamics, self organized systems, the kind of stuff that we can measure and [00:56:19] Speaker A: test and so on, so how the decomposition would work in other areas of, of cognitive sciences without grounding to the brain, and what comes to my mind is a large other area which is not cognitive science, but psychiatry. DSM 5, which is the popular and hated big book of my wife, who is a psychiatrist and many other psychiatrists, is that it's full with terms. And these terms were made in good faith by intelligent people and for A good purpose. So what is the strategy for decomposition? Unless we go back to the brain and from the brain perspective we will say this, two or three or four diseases are in fact has the same mechanism and perhaps the same substrate because they respond to the same treatment. For example, about 20 years ago, people started to, or psychiatrists started to use the drugs that have been used in epilepsy and took a long time to figure out or learn that oh, the reason why it happens is because many of those psychiatric problems are due to a hyperexcitability that if you had a recording from the right part of the brain would be classified by an epileptologist as seizure. So that's one example. So without going into the brain as a classifier and the final arbitrator, I don't know whether we can make a. We can make progress, of course, but in 2020, I think, you know, you are a prime example that you want to tie your work and research as much as possible to the brain rather than doing decomposition independent of the brain. [00:58:18] Speaker B: Again in a second. Okay, no, Paul, go ahead. You had a. [00:58:21] Speaker D: Well, no, I don't want to break the flow. Just on a personal note, I recently had Steve Grossberg on the show and one of the things that we talked about was his sort of cyclic method of developing his theories and his models. And it struck me as I've had it all wrong thus far. I'm slowly correcting myself as everything's a slow developmental process. But, but you know, I have thought, well, you know, like Yuri, I don't know that you think this way. But like I think, okay, we have all these brain mechanisms, like what could they do? And that's sort of a bottom up approach. And then I've also like sort of swam in the like, well, attention. You know what, how could that relate to consciousness? And how could these psychological terms relate and then how could they relate to brain? But Steve professes to, and I believe him because he's done it for a long time, he's been very productive. He begins with the behavior and he actually seeks out paradoxical behavioral findings and assimilates and synthesizes a large body of behavioral work from psychological experiments and psychophysics experiments. And then from there spends about five years, he says, thinking. And that leads to design principles that then he can implement in models and then start thinking about the mechanistic implementation of these things. And because he starts with behavior, it doesn't matter what the terms are for the psychological processes. Likewise, it really doesn't matter what the mechanisms are neuronally because his claim is that he does lets it stew for a long time and thinks about what sort of design principles could implement something like that. And in fact, that he originally started doing that before he knew anything about neuroscience and I presume about psychology, I [01:00:08] Speaker B: mean, I'm sympathetic to that. Right. So just to respond to, specific to Steve Grossberg's program with this particular part of his research, and in fact, that is sort of the core part of the paper that John Krakauer, Asif Ghazak, far out, Gomez Marin and Malcolm MacGyver and I wrote. That is to say there is a sort of core of identifying well characterized behaviors, hopefully behaviors that are really interesting to a creature, not like weird animal, not weird labs, that things that are just highly artificial, but really trying to identify, you know, reviving the ideas of ethology. And so not forgetting the incredibly important contributions from, you know, von Holst, von Frisch, Tinberg and Lorentz, who really were extremely careful at describing what might be going on, being inspired by that and turning that into experiments in neuroscience. That's. And so we are on record as saying, look, that aspect of extremely thoughtful characterization of behavioral should take an equal ground in neuroscience. It should be, should be revived as having this important role for stimulating research. So I think I would be in favor of that kind of view of growth research. [01:01:22] Speaker A: It's a very interesting thing. You just brought up ethology. I started out as an ethologist. I, I produced ethograms in cats and rats because that was the big thing at the time when the Nobel Prize was given to Timbergen, Frisch and Lawrence, who was, by the way, Heidegenberg's mentor. And the question in my mind is always that why did it die? Why did ethology die? And the reason is simple, because it never tied its systematic exploration to the brain. So here's one example. I spend enormous amount of time as a graduate student arguing back and forth with others whether Pavlovian conditioning is a similar substitution or it's an epistemological behavior, is the dog or cat or a rat is approaching the loudspeaker because it is so stupid that it thinks it's food or as the similar substitution predicts, or it is because it became a new goal as a result of the, of the contingencies. And this was the time when a term was introduced called auto shaping. Maybe you don't even remember what it is, but basically what it is is that you put an animal into a box and you present a lever, for example, and all the animal has to do is whenever the lever is presented, it just gets the food automatically. Doesn't have to behave except the lick. But animals can't help. Rats start pressing the lever. They shape themselves to do what Skinner shaped them to do. Pigeons pack on the light source and so on, because they can't help. This is how they are put together. That their brain is put together. They. That without behavior, without action, they learn. They never, ever get anything. They have to behave in order to receive something. So the debate went on and on and on. Whether one is the packing behavior or pressing behavior or approaching the loudspeaker is a consummatory behavior or a preparative behavior. And it was unending. I put an electrode in the hippocampus, and in an hour, the answer was clear. Because hippocampal theta oscillations are classifying all behaviors into two major categories, preparatory or consummatory. And there was always theta behavior. That was the end of the debate. So I realized in order to make progress, instead of doing ethograms, I should start poking the brain. [01:04:12] Speaker B: I mean, it's true that. I mean, it's an interesting story. I think you're right. It's interesting that. Why did ethology and neuroethology not just flourish suddenly? And maybe it was this focus on behavior, although the. [01:04:29] Speaker D: Wasn't it? Yeah, behaviorism in a way, but not [01:04:32] Speaker B: really, because behavior was. Prior to that. Behaviorism was killed off. [01:04:37] Speaker A: This was the 1970. [01:04:40] Speaker B: Behaviorism was effectively killed forever in 1959. [01:04:44] Speaker A: Exactly. [01:04:44] Speaker B: In 1959, it was over. [01:04:47] Speaker D: Well, is it. Because we thought we needed much higher control over the experiments. And so everything came into the lab and ethology was discarded because of that. [01:04:57] Speaker B: Maybe it's a good. I mean, so it does raise the interesting problem that we have currently in much of our research. Right. Which is this tension between experimental control and naturalism. Right. So on the one hand, the things we would like to understand are the sort of natural class of behaviors that organisms have. And then we bring people into the lab and have them, you know, press a button for a Gabor patch tilted to the right, which is, you know, like a little bit. But it's hard to do experiments and naturalistic preps. [01:05:24] Speaker A: Well, what killed behaviorism for me or Skinnerism is two things. One is the students of Skinner, Breland and Breland, who worked in. In Hawaii, have published a nice book called Misbehavior of Organism as a Response to Behavior of Organism, which showed that in many, many cases, animals Just cannot be trained, no matter how many trials you are trying to do. And that is left to Seligman's prepared, contra prepared and unprepared categories that there are certain rigidity in the brain or in evolution that doesn't allow you to train animals, for example, a rat to rear on its hind legs to avoid a shock. This is just naturally incompatible behavior. So the animal's brain is unprepared for that. And then later on learned helplessness and all the others came about. So this totality or globality of ask me any behavior and I can shape it, or I give the brain two different stimuli and the brain can associate anything you want was just not true. And it was very clear. So then the turn went back or the cycle turned back to the idea that you have to look at the brain. What are the constraints of the brain that allow things to happen and what things in the brain does not allow other things to happen? [01:06:51] Speaker B: Yeah, I mean, from the perspective of the psychological science and the Cognitive Sciences, Skinner's 1957 book Verbal Behavior was supposed to be the apex of explaining complex human function. And in 1959, Chomsky wrote a review of Verbal Behavior that was effectively a kind of just on logical grounds, an examination of could this work in principle on logical grounds? And it was a very. It's the ultimate reviewer number three. If you ever want to read something career ending, read Chomsky 1959. It's very neutral, it's very cold, clinical. And it also ended behaviorism and psychology that way. There was a one event for the cognitive sciences there. So that part didn't work. Interestingly, of course, with respect to AI, that's coming back. The notion of that Yuri criticizes and that I also relentlessly dislike of. This kind of tabula rasa approach is deeply ingrained in the AI approach to these problems. And I find it very surprising. It's going away a little bit, but it's sort of still. [01:07:54] Speaker A: Well, it's going away to bring in sleep as an offline helper and so on. But it's basically a blank slate. [01:08:03] Speaker D: No, but just. But there are priors in the form of the operations performed. So the classic example now is a convolutional neural network. Right. Well, it's still randomly initialized. [01:08:14] Speaker A: Exactly. It's still dead. [01:08:16] Speaker D: The weights are. But there's structure built into the network. I mean, I'm not saying it's there yet that AI is building in all of the things that, you know, it could, but it's moving in that direction. [01:08:28] Speaker A: So for the sake of argument, let's suppose for a moment that the idea that the brain's most important occupation or preoccupation is maintaining its own dynamic and everything comes secondary to that. If you take that, I would say, important message of my book, then where is the AI that starts out with that model where you say, oh, the first thing we have to do is to have a model that can have all the possibilities. It generates a realm of possibilities, what this system can absorb and can code for rather than the other way around. And I don't see it yet. [01:09:13] Speaker D: No, but we don't. But neuroscience doesn't know the right way how to construct a realm of possibilities. So it'd be hard for AI to code it in, correct? [01:09:24] Speaker A: Well, it's not. I would say we don't know. Yes, but we are making some progress at least. So, for example, my ex postdoc Ken Harris showed that the spontaneous patterns in the auditory cortex give you the realm of possibilities. So when you present signals of natural signals, artificial signals, they all fall into this realm of possibilities. There's no boundary violation. You cannot produce anything in the normal brain or in the human brain that is already not there as a dynamic. The most important thing for the brain is to keep that dynamic together. You know, there are so many things that the brain has to. The brain networks have to deal with and they are competing with each other, such as stability and reliance and perturbational sensitivity and so on. So the only way it can put together is to have this extraordinary diverse conglomeration of dynamic. And the moment you change it, then you have a disease. But learning, no matter how much we learn, the dynamic is not changing. So it is fundamentally different from the tabula rasa approach or the AI approach, where the complexity of the network scales with the amount of experience. My brain does not. Your brain does not on AI. You know, the shame or the problem is this bug called catastrophic interference. And there are various ways of dealing with it. Now, sleep as a saver is perhaps there, but catastrophic interference comes there because of the blank slate model. The brain solved that. And that's why I'm saying that I'd like to see somebody will be thinking about to start building a simple thing that can absorb only 500 items, but at least start out with the, with a thousand possibilities rather than start out with zero. [01:11:34] Speaker D: That's interesting. That harkens back to the original symbolic AI days where we were going to build in all the expert systems, where you build in all the knowledge. I know it's Very different. But it does remind me of that since we're on AI, because this is a topic that I wanted to talk about anyway. And Yuri, you just sort of made the case for the inside out approach to building AI, and my question was going to be whether any of this matters for AI. David, do you have a counter to that inside out approach to building AI? [01:12:04] Speaker B: No, I'm actually. You and I have talked about this before. I'm actually very sympathetic to Yuri's position on this. I think that's an AI that actually a kind of Neo AI that takes this position into account and evolution for that matter, and doesn't. It would be a very interesting new space. I'm super impressed by what AI systems can do. I think they're amazing and I think they're about engineering. I'm very moved. I'm impressed. I don't think they have anything to do with how minds and brains are organized. I enjoy reading about yet another network and yet another layer of convolutions and pooling. And I have the vaguest idea what it has to do with anything other than it solves an engineering question, but not a question about how biology works or, or about how minds work. [01:12:55] Speaker D: Is it intriguing to you though, that for instance, convolutional neural networks. And I don't mean to harp on this, it's just that that's where the state of the art is. [01:13:04] Speaker B: That's what your podcast is about. So you have to actually be. [01:13:07] Speaker D: No, well, I mean, I meant CNNs in particular, but that sort of structure explains a lot of the neurophysiological data seen in the ventral stream. [01:13:15] Speaker B: What do you mean with explains? It explains. So let's say the work of Danny ammons and Jim DeCarlo is an amazingly elegant way to show that you can get at some layer discrimination between visual images. I just don't think that that's what vision is about. [01:13:29] Speaker D: No, but when you look at the activity of the units, it matches fairly well with the activity of neurons. And so there's not a complete disconnect there. I mean, there's no. [01:13:40] Speaker B: No. Up to a certain level. No, I think so. This is actually an interesting conceptual point is up to a certain level of analysis, those seem like potentially descriptively adequate systems. The question is, are they explanatorily adequate so they capture good distinctions? That at some layer of, you know, the inferior temporal cortex, you can make discriminations between things. And yes, when you probe at some layer down, some layer of convolution, it looks a little bit like A receptive field. You can even make these cool super stimuli by exploiting the properties of these DNNs. I'll wait for the movie. You know, I'll wait for the movie. But at the moment I actually rather am in favor of the approach there that Yuri is arguing for, which is, let's take it into a very different direction, into a much more preformationist direction as we do know about intrinsic dynamics that are just there. And it would be extremely interesting to see what space that provides and what kind of AI. And yeah, I'm not a big fan of learning. I'm a. I'm probably way more nativist than you guys. But I'm impressed by what evolution and what genes carry and that they provide an operating system that's so rich that it provides you very clear boundary conditions of which, like the work of Ken Harris that Yuri mentioned. It's extremely interesting. Like you can't go outside of your coordinate system. And why would you. It would be very weird. I mean you have a really amazing operating system in the brain, Microsoft brain or whatever. It's a good. So that is going to be your. The thing within which you work. You don't give a trillion learning trials of some arbitrary scene. Biology is not the mother of all regressions. That's the daughter of evolution. [01:15:32] Speaker D: Okay, so before we go on because there are some specific claims that maybe we can go through here in a minute, but I want to ask what's at stake in this because does this issue. I'm stepping outside of AI and back to our psychology. High level, implementation level psychology, brain, inside out, outside in discussion. Does this matter issue in the long run because science won't science self correct either way and progress and work it out regardless of our opinions of how to proceed. Or. Or is this conversation more about speeding up that progress so that we can see it in our lifetimes? [01:16:13] Speaker A: So one level is practical. I can say that we have tried the outside in for a long time and we got to a point where we said the progress is slowing down. And so we can say why don't you try something else? And the inside out perhaps is an alternative that will induce new data, new ways of thinking and so on. And maybe we'll run into the same problem sooner or later. It's hard to predict. So whatever the reason, I think the time is right for now. There are deeper issues. The deeper issues is that the outside in inevitably produces concepts that are difficult to digest. In other words, in order to go from the perception of the world to action, you need some funny thing in the middle that you can call, you know, homunculus or consciousness or decision maker or anything. And that is very difficult to bypass. And as I try to illustrate it in my book, that the way how the outside in approach works is that you put the experimenter in a privileged situation because he's the one or she is the one who is presenting the world at the same time, recordings and interpreting the signals from the brain. So she or he has access to both, but the brain does not. The neurons that you are recording from have no clue what happens out there. That has to be grounded and grounding is the action. So you need a reader, you need a disciplined way of figuring out whether the correlation that you find is actually utilized by the brain. So in the lab, the. Not necessarily the mandatory attitude, but at least is the desire is that when we find a correlation, that we have to show that those patterns that we link to behavior actually are read by downstream readers. Now this brings me back a little bit to AI and especially robotics or the brain machine interface, paradoxically, or not the brain machine interface literature, or the people who are working on it subconsciously realize how important this is because they had a reader. This is, you know, the damn cursor. The cursor have to move up or down or somewhere. So there's no negotiation what you have to do, because every single thing that you feed in from the brain has to accomplish a goal. It has to be interpreted by the reader. And so this is the kind of attitude that said, if you go and start doing, building up the system from the point of. Of readers, then you are much better off than starting from outside and from a higher level and see how the brain responds and trying to interpret you as the experimenter of those responses without having a grounding mechanism. [01:19:31] Speaker D: So it's not the psychological process you're trying to. That's not the object that you're trying to build up a mechanism for. It is the action. [01:19:39] Speaker A: Exactly. [01:19:40] Speaker B: So I think. So your question was, what are the stakes? Does this matter? Or is this, you know. And there's two answers. One is this one is with Yuri. I agree there are practical concerns. I mean, what are the next steps to advance our agenda and asking one of the hardest questions in science? I think the stakes are unbelievably high. The stakes are what are. What is the parts list of the mind and brain? [01:20:09] Speaker D: Yeah, back to. [01:20:10] Speaker B: Yeah, that is the. What could be a more important list? I want to know, just like physicists want to know what are the elementary particles of how the universe is organized. I want to know what are the elementary particles of how minds and brains are organized. [01:20:22] Speaker D: But do you think there's risk that we won't get there or just that we'll get there much, much more slowly if we do it the wrong way? [01:20:29] Speaker B: Well, I'm not a. Let's say, is there a field called, you know, there's a field called history of science. I don't know if there's a futurism of science. I'm not a science futurist, I'm an optimist. I think that, you know, there is a fact of the matter. I'm a naive realist. I believe the world is as it is and we can figure it out. And I would like to use the best, put our best foot forward. Right. And so I think the slightly too casual dismissal of the contribution of the psychological sciences comes at a cost that I don't want to incur. I mean, that is the main argument of the. Of the paper that Federico and I wrote in response to Yuri's paper and book. That is, it's too quickly dismissing the conceptual architecture of cognition because, of course, absolutely true, there's a lot of lame and shitty work that's legitimate and it's annoying. But there's also progress and good work. And we shouldn't, as we said, don't throw the baby out with the bad science bath water. That is, these are realistic, well substantiated hypotheses about elementary. The elementary particles of the mind and brain. And they should have the same kind of status as our hypotheses about whether the neuron doctrine is right or whether everything, you know, should be situated at the level of, you know, a dendritic spine or whether we should be looking in genes. Let me make one more. So I think the stakes are really high and really exciting. Important. We want to know how the basics work in the most fascinating field. One thing I want to disagree with, or partially disagree with Yuri, is that the arbiter of everything is action. I think that's certainly part of it, but I would submit that one intermediate step, and maybe a prior step to that is actually storage. As one of the remarkable things we have, you know, in this sort of pre configured or preformed nervous system operating system is the ability to store stuff. Some of it is already there and some of it we have to put in. And that forms the basis for many things, sometimes including action. So I think one of the deepest questions that we owe in the neurosciences and the cognitive sciences is how does storage actually work? You know, it's extremely difficult. And I think that the answers we have at the moment strike me as very unsatisfying. Like oh, it's the pattern of synaptic forms or something like that. That's, you know, maybe or maybe not, but it seems not. So I think that's one of the things for the future to dig in deep because it's going to lie at the basis of the links between how brains work and how minds work, which I take to be more or less the same thing. Right, so they're different coins of the inquiry. [01:23:11] Speaker A: Well, we work on memory, but memory and storage is totally useless without action. There's no need, there's no. If something in the brain is not implemented and the implementation can happen only through action, then it doesn't exist. [01:23:28] Speaker B: But is action for you a larger. Just a word that also includes thought. [01:23:31] Speaker A: Thought. [01:23:32] Speaker B: Okay, so that's what I mean. So it's a larger thought. [01:23:36] Speaker A: What I call is a deferred action. [01:23:38] Speaker D: I was going to ask if memory is a deferred action. [01:23:40] Speaker A: Memory is a deferred action too. So you know, what is the largest. Take the larger agenda to me with the inside out versus outside in that if we build our logical system from the inside out, maybe the middle will disappear and those problems that are large baggage of the outside in, such as the little man, such as decision making and all these very difficult concepts will be addressed in a totally different light. [01:24:12] Speaker B: Do you have an example how that would work? [01:24:14] Speaker A: Well, the example is the way how I view the whole thing differently. So in chapter three I discussed at length corollary discharge, the difference between perception action and action perception. So this asymmetry occurs because every single time the brain sends an output to the actuators or a thought, it always signals back to what we call sensory part. And this is a. This is, this is the key moment. Because once the brain figures out and learns how to disengage from the muscles, the same mechanisms can be used for signaling about what if scenarios, what happens, what would happen if I would have made this action without actually doing that. And this is called thinking. So starting from the action and going back way. [01:25:17] Speaker B: But what are the elements of that? I mean I'm all on board with corollary discharge. And in fact it's very close, although not identical cousin that you have internal forward models that have not just corollary [01:25:29] Speaker A: discharge, but internal forward model is a variation of the corollary. Corollary discharge is a down to earth mechanism. [01:25:37] Speaker B: But what are the elements of that. Right. So when that becomes thought, I mean, so what are the conceptual or representational primitives of that corollary discharge? [01:25:47] Speaker A: So we can start with evolution and look how this works in the cricket. And then we see how this basic mechanism is being. Getting larger and larger and larger. And I give several examples in my book. Is that even emotions explainable by coronary discharge. I have a chapter that you. Not a chapter, but a paragraph at least there, that you didn't criticize. That language is probably based on corollary discharge. And the thought itself, the thought process is based on corollary discharge or this mechanism or something like that. Mechanism. [01:26:28] Speaker B: Yeah. I mean, the reason I didn't criticize that particular. So I'm actually very much in a huge. Well, it's not a notion of fandom. I think that there's something extremely useful about that concept because I myself have done a series of experiments actually on how corollary discharge plays a key role in speech production. The question whether it scales up and there. I know what the primitives are. The question is whether it scales up or scale sideways to other aspects of thought. Right. And that's a more difficult. [01:26:56] Speaker A: Yeah, I agree with that. But this is what I try to scale up as an existing mechanism, as something that inherited from phylogeny and that can be used and exploited in multiple ways. [01:27:08] Speaker B: Yeah, no, that's an example of where I think there's a really cool mechanism. I think for the first time, perhaps I've seen in the work of Ernst Mach, the physicist, talking about that because of eye movement control. So you have to have a copy sent into the future in a quite explicit paragraph in 1880s or something like that. And then later in the work of von Holst, for instance, I just want [01:27:35] Speaker D: to take a time out and make sure. Because we're coming to time, I suppose I just want to make sure there's not something that you want to make sure that we get to. [01:27:44] Speaker B: But I want to find. While Yuri says something intelligent, I want to find a quote to really nail him to the. [01:27:52] Speaker D: Okay. Oh, he's pulling out the book. [01:27:54] Speaker B: I took some notes here to really, you know. [01:27:57] Speaker D: Did you want to respond to that, Yuri? [01:27:59] Speaker A: Well, I'd like to. I plan to say something and I'd like to say it now is that it's an interesting thing that we have this debate and about the origin of these words. And you may wonder why this is so natural that we have this tendency or this urge to come up with words and terms and the way I try to formulate, this is something like this, that in every animal, semantic information is derived from personal experiences, from episodic memories. The first time you see a dog is an episode. It's your personal experience. But when you come across many dogs, the spatial temporal conditions of those specific episodes are stripped off, and what is left is the semantic entity of the dog. So learning semantic knowledge typically develops from ego specific individual interactions. This applies to all animals. However, and this is a funny part that humans, this process is cut short. We have externalized, not we evolution, externalized a great deal of brain functions through artifacts and especially through language. So now we can learn explicit semantic knowledge by absorption, not by individual episodic memory. That is learning from others who name things for us. We learn about giraffes, tigers, also about angels, Santa Claus and things like Covid, the NASDAQ index, and galaxies. And you can go on now. Where does this lead to us? This is a tremendous advantage, but it comes with a cost that we accept their existence, the existence of these terms, without ever experiencing or questioning them. When these terms are used, especially by authorities, they become real. But you may wonder, are they really real? It is so easy to say that the reason we understand something is because the brain processes information. Or we can say we sense time and space because we have place cells and time cells in the hippocampus. So these are the statements I have problem with, and I illustrate it in various ways. And I have a cartoon at the beginning. I think in chapter two that the prefrontal cortex has already at least 200 terms that have been associated one way or another with some terms. So the reason why we tend to name things is because we have this urge, because we think that once we have a word for it, that explains. So the crux of the problem is that we use words and the terms to explains, to explain things that have to be explained. This is what I call the. The mixing of the exponent and exponents. And this is, I think, something that David picked on also. And there is, I think, a fundamental agreement that the two aspects shouldn't be used interchangeably. But how we get away from it is the big issue. [01:31:16] Speaker D: But isn't part of your agenda to create new vocabulary based on neural principles, new grounded vocabulary? [01:31:22] Speaker A: I don't necessarily need new words. If for whatever reason a new word will clarify things better, let's go for it. But I'm happy to work with the word interneuron that was created a long time ago, even though now we know that many garbage neurons have Long projection. But when we refer to interneurons, we already defined it. We know exactly what we are talking about. [01:31:47] Speaker D: I'm no language expert, but we do have at least one language expert among us. The way that new words are created is not. Someone doesn't decide to create a new word and then it becomes a new word. It's someone, it gets used and then something sticks. [01:32:04] Speaker B: But I think, I mean, what Yuri is saying, we don't need to invent new words. We can, but it's the scope of the concepts that a word, it's the reference or what is the particular meaning of a word that's at stake when we use it in the scientific context. I mean, so I think with respect to the explanation question, so I want to just narrowly focus on the question of the relationship between cognition and the brain. If that is the focus and one of the core parts of neuroscience, then I guess Yuri points out very early in his book, which I have underlined right here, the, the thing to be explained should be the activities of the brain, not the invented terms. And that's wherein I think we. That's the implementation of chauvinism. That is, I think what is to be explained is, as importantly is actually the invented terms. On the view that the invented terms are just as carefully submitted to scientific analysis, research, empirical verification. That is, the terms that we try to understand, even something as offensive as attention can be extremely, you know, can be subjected to the same level of scientific scrutiny and empirical scrutiny as biology. And they are ultimate. If we're trying to understand aspects of mental function, then those are the things that are the explananda as well, not just the activities of the brain. And that is, I think, where we. So I would submit that you, you must still have some of, I don't know if we should call it outside in, but that the. That level of description needs to be sustained, just as the level of description of the neuroscience needs to be sustained. And I think that. And that the arrow of information cannot just flow from the characterization of the activity of the brain in one direction, of the activity of, you know, psychology or cognition that I would feel strongly about. [01:34:15] Speaker D: So Yuri mentioned that his hope and he's optimistic that the middle will disappear. Maybe you both, I don't know if you both want to comment, but is there a middle? And if so, will it disappear or will it disappear? And if there's not a middle, I suppose that means that the emergent properties that are implied and described when we use psychological terms will forever be separate from the Emergent properties that we describe when we use implementation level terms. So is there a middle or will it disappear? [01:34:51] Speaker B: I'm not sure I understand what the middle is at the moment, actually. [01:34:56] Speaker D: Well, so [01:34:58] Speaker A: the notion of the middle is that in order to go from the outside in and to translate what the world is telling the brain to action, that has to be. The logic goes like this. Brains are created to learn about the world. There are things out there that are good or bad, you have to evaluate and you have to choose good. And once you make the decision, you can act. So this is this serial process which you can call perception, decision making and action cycle. I put it the other way around. I said there is a action perception cycle. There is no perception without action. That's a big claim. And I try to make it clear why I made it. So the middle is there because it has names. Question is whether those names really refer to something or not. And is the decision making process is a vague idea and that is present and omnipotent in a single neuron. Because an action potential, you can conceive it as a decision. And if you make it so ubiquitous, then it's no longer capturing those things that philosophers and maybe neuroscientists were talking about. So this is what I'm saying, that with that kind of thing it may go away. And I hope that if we approach the brain from inside out, we may build up a different conceptual system, perhaps with different words and different problems to attack than the traditional outside in. [01:36:43] Speaker B: I mean, I guess what you owe then is a sort of. I mean, this is a way to answer Jon Krakauer's question, right? I mean, what you would owe is to say, I have a new conceptualization, maybe with new or modulated vocabulary that actually can be an account of some mentalist phenomenon. That's John Krakauer's challenge, right? I mean, how would you, using the infrastructure or conceptual structure or vocabulary of the neurosciences, then account for the psychological phenomenon? So you could tell a story that way, but you still need to sort of figure out what is that chain of argumentation. [01:37:21] Speaker D: In that case, I have a feeling that John Krakauer would, even if that were to occur, he would maintain that the term the psychological processes are on a different level and the vocabulary used are describing different properties of the phenomenon. [01:37:37] Speaker B: Again, are we talking about the scientific use of the terminology or the informal use of the vocabulary? We will not go away from that because they are psychologically useful. Concepts like imagination, greed, will, self are useful. That's how we attribute that's sort of the modern computational theory of mind. That is your belief system, right? Or knowledge system is that kind of terminology. And if you get rid of that, what will you put in its place? To have a conversation about how we're, you know, about how we interact and act. [01:38:14] Speaker A: Exactly, exactly. We cannot have a conversation without words. [01:38:17] Speaker B: Without words. [01:38:18] Speaker D: You don't want to get rid of it, right, Yuri? You don't want to get rid of it. [01:38:21] Speaker A: We need two things. One is to realize that these are conversational pieces. The second one is to try to justify them. The problem David tried to criticize has been criticized even more so by my psychiatrist wife that, well, how do you derive from your sodium channels and spines ever depression or even tinnitus? I had a wonderful experience a couple of years ago. A rich person from Monte Carlo invited a group of tinnitus experts and neuroscientists who had no knowledge whatsoever about tinnitus. And we were imprisoned together for a week. And we had to talk about how to solve this problem. And, you know, when you comfort yourself with something like this or depression or schizophrenia and said, well, how many years of my approach at this elementary level would take to get an answer? And of course, you know, I can, I have to tell my wife that, you know, we keep conversation about words and modify it as these symptoms. And so I allowed them. Don't wait for neuroscience, because then you have to wait for a long period of time. But it doesn't mean that's hopeless. We are working very hard. We are making inroads in various issues, and perhaps we can build up someday a vocabulary or a symptomatology or a drug discovery programs based on brain rhythms. Every single psychiatric disease is associated with a different constellation of problems, of rhythms. So what if I can find a drug that. Or work on a drug that works on the thalamocortical alpha or another one that affects sleep spindles in certain segments of the brain. These are neuroscience questions. And all of a sudden that kind of drug treatment based on the treatment of the rhythms will fix, let's say, one disease or several diseases simultaneously. So this is the kind of far reaching consequences of the Inside out program. [01:40:42] Speaker B: Is it really inside out? I just don't see that. I just don't. I disagree with you on that. I mean, because you started this with the very interesting point that, you know, [01:40:49] Speaker A: well, inside out in a way that you take the brain as your primary target and say, oh, let's use the neural mechanism as a testbed to invent a new drug, right? [01:41:01] Speaker B: But you're finding. I mean, what you're saying is, you know, the really, the alluring possibility is that you find a taxonomy of, for instance, in this case, let's say, psychiatric disorders that have shared underlying things, let's say because of rhythm, changes in brain rhythms or dysrhythmias or arrhythmias. But that's because of the presupposition that there would be underlyingly, conceptually similarities in the things ultimately underlying the pathology. I mean, that's a very strong theoretical position that you're bringing to that task. I mean, that's as outside in as it gets. I mean, you have a clear theory. [01:41:38] Speaker A: No, the outside in gives you DSM 5. DSM 5 is full with a lot of things. What I'm saying is that. That DSM 5 could be grounded by brain mechanisms and maybe simplified much more effectively than putting 500 psychiatrists above the table and negotiate. [01:42:02] Speaker B: That must be true for many reasons. I mean, just to push your button on one more thing, just because I was looking through some of my notes and that came to my mind. So there's in your a chapter that for me was very important, was just the paper, the chapter on space and time. I think it's very near and dear to concerns I have, you say somewhere, which I think is very important because it's also germane to my own particular research interests where I was paying close attention. You say the hippocampal system may be responsible for constructing sequences of information chunks while chunk content is encoded and retrieved from the neocortex. And that seems to me a very reasonable hypothesis and worth, you know, digging deep into. The critical question for me is the word content. What is the chunk content? And how would you actually ask that question and go about it from. So I have a clear way of how I would go about it from the putative outside in way. But I want to know how you would go about it because it's a little bit mystifying to me. [01:43:09] Speaker A: Well, the content is a short list of things that could be simplified in our world by the animal running through a maze. And item one is corridor one or the beginning of corridor one. And the second one is somewhere else and so on. And these segments are concatenated into a sequential order of firing of the neurons. So when I have a sequential activity, the question is whether those neurons actually represent space or they are pointers to the neocortex, where the world is represented or is mapped. And the hippocampal sequence just helps to link together these contents into an episode. So this is an example where I would say it's experimentally relatively easy to put your finger on what is the potential contact, because that could be measured in centimeters or it could be measured or exemplified with corridors. I understand what you're asking is, when it comes to a more complicated issue, then you would like to know how to break down the content. And my answer is always that, you know, you have to have a simple situation, such as an animal running through corridors, and then we can go from there. [01:44:38] Speaker B: Yeah, I mean, I. So I certainly see you have a way of operationalizing the question for a very clear experiment in this case, sort of an animal prep. And I guess the kind of. You're right to point out I would be interested in what content means in a. In a more. In a way of what I have actually as stored content. Right. So I have a. I have a mental representation of the episode of this discussion. And I know that, you know, you were here and Paul's here and I'm here, and these are chunks of information, presumably neocortical. And I want to know, well, how are those things done? And I can. Then you have a story for how to actually connect those through hippocampal mechanisms into sequences. We start talked about this, we talked about that, we talked about the other thing, but for me, the CASOs, KNEXOs, the crucial thing is actually the original instantiation of the primitives of this thing like that I have. So this is what's very, very difficult [01:45:33] Speaker A: you raise in the. I forget which paper you had, this review paper that you wrote alone about the problem of alignment and mapping. [01:45:41] Speaker B: Yes, that's right. [01:45:42] Speaker A: And indeed, this is an interesting thing because mapping, of course, is the big preoccupation in the hippocampal research. And the hippocampal map assumes that somehow space is mapped onto a hippocampal structure and its mechanisms to serve to represent space. The problem with this approach is that space cannot exert any effect, any effect on anything, because space is a concept only things can affect. In physics, we have no space sensors in our bodies or in the brain. No matter how attractive it is to explain hippocampal function in terms of space, in reality, we are examining relationships between things and objects, and for convenience, we refer to them as space. But things are not in space. Space is the things themselves. Now, this is something that we debate among ourselves, but I think it somehow relates to what you are saying about the relationship between the primitives and the big concepts of language, because the big concepts are making up language. Language is not this, but it is the primitives themselves in a particular constellation. And so when you are taking it apart and filling it with new content and tying it to bray mechanism, this is what you call alignment. And it resonates with me as well. And I think we already clarified that. I don't think that you have to, you know, you go blindly in a physiological experiment and then automatically something will come out. But you go in with a lot of disciplined or naive primitives and this is what you are trying to find, whether it matches the brain mechanisms or not. [01:47:38] Speaker B: Yeah, I mean this is sort of what Federico and I would say. Well, ultimately the day to day practice of these experiments is an abductive process. That is, you have some kind of intuitive notion of what I'm looking for, then you really go into the nitty gritty, then you refine the concepts because you sort of, you don't do it purely inductively because that has just never worked. [01:47:58] Speaker D: Exactly. [01:47:58] Speaker A: So this is what I said, that there are two major parts of the brain's organization. One is what I call the good enough brain. And the good enough brain is there for us to do what you just said that under all conditions we have an answer. There is always a preconception. The brain always knows the answer. I cannot show you anything in this world that you would say it doesn't exist. Because you will automatically say, oh, this is something like your brain immediately generalizes, it interprets it. The brain always interprets. And this interpretation process is of course what goes into the everyday business. When we go into the lab and do experiments that we interpret it and then it requires a particular eye who sees it differently than others that aha, this is a little bit different than what we are used to. And you know, this is not a heroica, but Feynman would say the most important moment in the scientist's life when he says, hmm, it's funny, I thought that was Asimov. [01:49:03] Speaker D: Was that Feynman? [01:49:04] Speaker A: I thought it was Feynman from the book of professor you are joking or something like that. [01:49:10] Speaker D: He might have been quoting Asimov. I don't know, [01:49:14] Speaker A: I don't have a stake in it. But I think I agree with the principle that indeed that's what we do because that's the discovery process that all of a sudden our preconceptions are confronted with something. And this is, I think what mainstream neuroscience should do now is taking those concepts that we inherited and confront them and see if we get some surprises here and there. [01:49:38] Speaker B: I mean, if you didn't have a preconception going in. You would never have the surprise moment. It's because you approach it with a preconception. You say, huh, I wonder what that's all about. Right? Otherwise it would be just description, description, description, stamp collecting. It would just be step core. [01:49:55] Speaker A: That also goes in neuroscience these days. [01:49:58] Speaker B: Yes. [01:49:59] Speaker D: Yeah. I wanted to spend more time on abduction because I think I really enjoyed that later piece in your article, David, but we gotta go. And this is always the risk you run when you have a couple people who are almost as smart as Franz Josef gall, but not quite. You know, we could clear. We could just go on and. Anyway, but I hope that we move the needle and at least gave people food for thought. And you know, of course I'll point people to both your papers, that, that series of articles and the introductory article and, and Yuri, your book, which is again, wonderfully done. So thank you for the book and thanks for the article. [01:50:39] Speaker B: I mean, I do think, you know, like I said at the very beginning, I think it's. I hope you can say something where. But the just, it's really just unbelievably useful to have debates go back and forth. Right. I mean even this is because, you know, both Yuri had the experience. I've had the experience with students and trainees who just. They are uncomfortable about debate now. And we. And it's exactly the opposite that you should have. [01:51:00] Speaker D: Right. [01:51:00] Speaker B: To sort of say, well, I just, I'm just not understanding. I need to look at it from different. And it's kind of fun, it's engaging. And I think it's part of our sort of professorial responsibility. Say that we get joy from that. [01:51:14] Speaker D: But people debate in a different way these days. It seems like more name calling and definitive statements rather than. I think this, therefore that it is. You are this. Because that. So it's a. [01:51:26] Speaker B: Well, that's okay, but that's kind of lame. [01:51:28] Speaker D: It's lame, but I think that that might be a reason why people are less. I don't know. I don't really know. [01:51:35] Speaker A: The other thing I think is the lack of time discussing a paper for me and for my group. 30 years ago it took a day. Now we are discussing two papers in a lab meeting. So here is the deal, David. We will pick a paper of yours and we will invite you to present it. Yes, that's a great idea. My group is a little more vicious than the average group and they will. But they're probably like, take your part. [01:52:10] Speaker B: They're probably like my group. My lab meetings get very vigorous. It's fun. No, that's exactly. I would love to. [01:52:18] Speaker D: So, anyway, guys, I'm sorry we ran out of time. Thank you so much for the fascinating conversation. And, you know, hopefully, maybe we can do it again sometime. Who knows? [01:52:25] Speaker A: Thank you so much for organizing this, and thank you for the outside comments. And thank you, David, for your effort and your kindness and your criticism. [01:52:34] Speaker B: Thanks. Thanks, Yuri, for tolerating me. Thanks, Paul, for organizing it, and I look forward always to more. It's great fun and it's important.